Abstract: The emergency assessment of acute suicidal risk in adolescents is a daunting clinical challenge because our current ability to predict suicide attempts is weak, and because the risk for suicide attempts in suicidal adolescents is high. Nevertheless, there have been no studies that have examined the best approaches to the prediction of suicidal behavior in suicidal youth presenting to a psychiatric emergency department (PED). To address this research gap, we propose a study of 1800 youth presented to a regional PED, 1350 of whom present for evaluation of suicidal risk, in which youth are assessed in the PED, and followed up at 1, 3, and 6 months to determine which youth have made a suicide attempt. We propose 3 complementary approaches to assessment of suicidal risk. First, in this competitive renewal, we build on our success in developing computerized adaptive tests for 6 diagnostic groups, plus suicidal risk, during our previous project period. These self- and parent-reports can be completed in a total of 10-15 minutes. Second, because theory-driven assessments of suicide risk have strong predictive power in adults, but have never been tested prospectively in adolescents, we propose to test the predictive power of measures of Shneidman?s psychache (mental pain) and Joiner?s Interpersonal Theory of Suicide, which posits interactive roles of perceived burdensomeness, thwarted belonging, and acquired capacity for suicide in driving suicidal risk. Third, we aim to use machine learning (ML) and natural language processing (NLP) of electronic health records (EHRs) to identify youth at risk for suicide attempts. We hypothesize that each of these approaches: (1) CATs for suicide risk and for depression, anxiety, bipolar, ADHD, oppositional defiant, and conduct disorders); (2) theory-derived measures of suicidal risk; and (3) ML and NLP of EHRs, will each be superior to clinical assessment alone in the prediction of attempts, and that the combination of the 3 approaches will be more powerful than any one of these approaches alone. This study is innovative because it is one of the first to use CATs for the prediction of suicidal risk, in a consistently high risk population, the first prospective test of two leading theories of suicide in adolescents, the first to use machine learning and natural language processing to identify EHR predictors of suicide attempts in adolescents, and the first to test a combination of approaches to the identification of imminent suicidal risk in adolescents in a sufficiently large, high risk sample. The study is of potentially high impact because it could identify brief, easily disseminated assessment strategies to identify youth at high risk for suicidal behavior and add to clinicians? ability to match intensity and type of resources to those at greatest clinical need. The approaches to be tested in this study could yield assessments that reflect the two imperatives of emergency mental health care: brevity and accuracy. With better ability to identify who is at risk for suicidal behavior, we will be in a much stronger position to identify who needs intervention and reverse the disturbing, decade-long trend of increases in adolescent suicide and suicidal behavior.